Proceedings of the 14th International Conference on Parallel Architectures and Compilation Techniques
An Application of Constraint Programming to Superblock Instruction Scheduling
CP '08 Proceedings of the 14th international conference on Principles and Practice of Constraint Programming
Constraint-Based register allocation and instruction scheduling
CP'12 Proceedings of the 18th international conference on Principles and Practice of Constraint Programming
| Hi-index | 0.00 |
HP and Intel's Itanium Processor Family (IPF) isconsidered as one of the most challenging processorarchitectures to generate code for.During global instructionscheduling, the compiler must balance the useof strongly interdependent techniques like code motion,speculation and prediction.A too conservative applicationof these features can lead to empty executionslots, contrary to the EPIC philosophy.But overuse cancause resource shortage which spoils the benefit.We tackle this problem using integer linear programming(ILP), a proven standard optimization method.Our ILP model comprises global, partial-ready code motionwith automated generation of compensation codeas well as vital predication.The ILP approach can - withsome restrictions - resolve the interdependences betweenthese decisions and deliver the global optimum.This promises a speedup for compute-intensive applicationsas well as some theoretically funded insightsinto the potential of the architecture.Experiments with several hot functions from theSPEC benchmarks show substantial improvements:Our postpass optimizer reduces the schedule lengthsproduced by Intel's compiler by about 20-40%.The resultingspeedup of the routines is 16% on average.